pipette
ESEspañol

Algorithm-aware Statistical Inference in High-dimensional Estimating Equations

Tongyu Li, Alexander Giessing

Preprint

In the authors' words

Iterative algorithms are integral to many statistical procedures, whereas sampling uncertainty along their iteration is generally not quantified. We develop simultaneous inference for discrete trajectories generated from high-dimensional estimating equations, with iteration-specific population estimands that preserve algorithmic regularization. Our proposed fixed-Jacobian bootstrap propagates perturbations along the observed trajectory without repeatedly running the original nonlinear algorithm. We establish uniform first-order expansions and bootstrap validity over possibly infinite sets of contrasts and iterations under stability, entropy and moment conditions. Simulations in sparse linear regression, paired Fisher discrimination, and temporal-difference learning demonstrate accurate simultaneous coverage. A micro-randomized trial application illustrates inference for a treatment trajectory with participant-level clustering.

Main resultThe abstract does not state a limitation.

Appeared: Tuesday, November 10. arXiv. Preprint, not yet peer-reviewed.